In [6]:
fig,ax=plt.subplots(nrows=4,ncols=4,figsize=(20,20))
ax=ax.flatten()
for i,j in enumerate(train_images[:16]):
ax[i].imshow(j) #already in minimal scaled range
ax[i].title.set_text(train_labels[i])
import numpy as np
import tensorflow as tf
import os
import matplotlib.pyplot as plt
import cv2
import numpy as np
from sklearn.model_selection import train_test_split
import warnings
warnings.filterwarnings("ignore")
WARNING:tensorflow:From D:\python\lib\site-packages\keras\src\losses.py:2976: The name tf.losses.sparse_softmax_cross_entropy is deprecated. Please use tf.compat.v1.losses.sparse_softmax_cross_entropy instead.
#PATH ALLOCATION
data_dir=os.path.join('data')
classes=os.listdir(data_dir)
classes
['Africanized Honey Bees (Killer Bees)', 'Anaphe reticulata', 'Anomis sabulifera', 'Ants', 'Aphids', 'Apion corchori', 'Armyworms', 'Beet Armyworm', 'Beetle', 'Black Hairy', 'Bollworm', 'Brown Marmorated Stink Bugs', 'Cabbage Loopers', 'Catterpillar', 'Citrus Canker', 'Colorado Potato Beetles', 'Corn borer', 'Corn Earworms', 'Cutworm', 'Earwig', 'Fall Armyworms', 'Field Cricket', 'Flea beetle', 'Fruit Flies', 'Gall midge', 'Grasshopper', 'Grub', 'Hornworms', 'Jute Hairy', 'Leaf Beetle', 'Leafhopper', 'Mealybug', 'Miridae', 'Mosquito', 'Moth', 'Pod Borer', 'Red Mite', 'Sawfly', 'Scopula Emissaria', 'Slug', 'Snail', 'Spider Mites', 'Stem maggot', 'Stem_borer', 'Termite', 'Termite odontotermes', 'Thrips', 'Wasp', 'Weevil', 'Western Corn Rootworms', 'Wireworm', 'Yellow Mite']
#REMOVING DOGGY IMAGES
import cv2
import imghdr #cv2 uses BGR,matplotlib uses RGB
img_ext=['jpeg','jpg','png','bmp']
for img_c in classes:
for img_v in os.listdir(os.path.join('data',img_c)):
img_p=os.path.join('data',img_c,img_v)
try:
img_mat=cv2.imread(img_p)
img_tip=imghdr.what(img_p)
if(img_tip not in img_ext):
print('Image not in ext list {}'.format(img_p))
os.remove(img_p)
except Exception as e:
#print('Issue with image {}'.format(img_p))
print(e)
# Load and preprocess data from subfolders
def load_data(folder):
images = []
labels = []
class_names = sorted(os.listdir(folder))
for class_index, class_name in enumerate(class_names):
class_path = os.path.join(folder, class_name)
for image_name in os.listdir(class_path):
image_path = os.path.join(class_path, image_name)
image = cv2.imread(image_path)
# Convert the color space from BGR to RGB
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
image = cv2.resize(image, (256,256))
images.append(image)
labels.append(class_index)
return images, labels
# Load and preprocess image data
images, labels = load_data('data')
# Split data into train and test sets
train_images, val_images, train_labels, val_labels = train_test_split(
images, labels, test_size=0.2, random_state=42, stratify=labels
)
# Convert lists to numpy arrays
train_images = np.array(train_images)
train_labels = np.array(train_labels)
val_images = np.array(val_images)
val_labels = np.array(val_labels)
# Normalize pixel values to the range [0, 1]
train_images = train_images.astype('float32') / 255.0
val_images = val_images.astype('float32') / 255.0
# Print shapes for verification
print("Train images shape:", train_images.shape)
print("Train labels shape:", train_labels.shape)
print("Validation images shape:", val_images.shape)
print("Validation labels shape:",val_labels.shape)
Train images shape: (4097, 256, 256, 3) Train labels shape: (4097,) Validation images shape: (1025, 256, 256, 3) Validation labels shape: (1025,)
fig,ax=plt.subplots(nrows=4,ncols=4,figsize=(20,20))
ax=ax.flatten()
for i,j in enumerate(train_images[:16]):
ax[i].imshow(j) #already in minimal scaled range
ax[i].title.set_text(train_labels[i])
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D, MaxPooling2D,Dense,Flatten,Dropout
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.layers import Activation, Dropout
from tensorflow.keras.callbacks import TensorBoard, EarlyStopping
model=Sequential()
model.add(Conv2D(16, (3,3), 1, activation='relu', input_shape=(256,256,3)))
model.add(MaxPooling2D())
model.add(Conv2D(32, (3,3), 1, activation='relu'))
model.add(MaxPooling2D())
model.add(Conv2D(16, (3,3), 1, activation='relu'))
model.add(MaxPooling2D())
model.add(Flatten())
model.add(Dense(256, activation='relu'))
model.add(Dropout(0.1)) # Adding dropout for regularization
model.add(Dense(128, activation='relu')) # Adding another dense layer
model.add(Dropout(0.1)) # Adding dropout for regularization
model.add(Dense(52, activation='softmax'))
# Compile the model
# Define a custom optimizer with a smaller initial learning rate
#custom_optimizer = Adam(learning_rate=0.01)
#model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
# Print the summary of the model architecture
model.summary()
WARNING:tensorflow:From D:\python\lib\site-packages\keras\src\optimizers\__init__.py:309: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.
Model: "sequential_2"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_6 (Conv2D) (None, 254, 254, 16) 448
max_pooling2d_6 (MaxPoolin (None, 127, 127, 16) 0
g2D)
conv2d_7 (Conv2D) (None, 125, 125, 32) 4640
max_pooling2d_7 (MaxPoolin (None, 62, 62, 32) 0
g2D)
conv2d_8 (Conv2D) (None, 60, 60, 16) 4624
max_pooling2d_8 (MaxPoolin (None, 30, 30, 16) 0
g2D)
flatten_2 (Flatten) (None, 14400) 0
dense_6 (Dense) (None, 256) 3686656
dropout_4 (Dropout) (None, 256) 0
dense_7 (Dense) (None, 128) 32896
dropout_5 (Dropout) (None, 128) 0
dense_8 (Dense) (None, 52) 6708
=================================================================
Total params: 3735972 (14.25 MB)
Trainable params: 3735972 (14.25 MB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________
tensorboard_callback = tf.keras.callbacks.TensorBoard(log_dir='logs')
history = model.fit(train_images, train_labels, epochs=20, validation_data=(val_images, val_labels), callbacks=[tensorboard_callback])
Epoch 1/20 129/129 [==============================] - 53s 335ms/step - loss: 3.9486 - accuracy: 0.0271 - val_loss: 3.9288 - val_accuracy: 0.0517 Epoch 2/20 129/129 [==============================] - 36s 274ms/step - loss: 3.7878 - accuracy: 0.0703 - val_loss: 3.6690 - val_accuracy: 0.1034 Epoch 3/20 129/129 [==============================] - 40s 314ms/step - loss: 3.1220 - accuracy: 0.2287 - val_loss: 3.5227 - val_accuracy: 0.1688 Epoch 4/20 129/129 [==============================] - 34s 262ms/step - loss: 1.9319 - accuracy: 0.4945 - val_loss: 4.1518 - val_accuracy: 0.1922 Epoch 5/20 129/129 [==============================] - 40s 313ms/step - loss: 1.0306 - accuracy: 0.7366 - val_loss: 4.9104 - val_accuracy: 0.1971 Epoch 6/20 129/129 [==============================] - 37s 291ms/step - loss: 0.5366 - accuracy: 0.8640 - val_loss: 6.2540 - val_accuracy: 0.1922 Epoch 7/20 129/129 [==============================] - 38s 292ms/step - loss: 0.3897 - accuracy: 0.8990 - val_loss: 6.8553 - val_accuracy: 0.2000 Epoch 8/20 129/129 [==============================] - 40s 313ms/step - loss: 0.3035 - accuracy: 0.9226 - val_loss: 6.7754 - val_accuracy: 0.1873 Epoch 9/20 129/129 [==============================] - 32s 252ms/step - loss: 0.2828 - accuracy: 0.9297 - val_loss: 7.0596 - val_accuracy: 0.1912 Epoch 10/20 129/129 [==============================] - 32s 244ms/step - loss: 0.2052 - accuracy: 0.9522 - val_loss: 7.7635 - val_accuracy: 0.1961 Epoch 11/20 129/129 [==============================] - 36s 277ms/step - loss: 0.1547 - accuracy: 0.9573 - val_loss: 7.7312 - val_accuracy: 0.1922 Epoch 12/20 129/129 [==============================] - 41s 315ms/step - loss: 0.1338 - accuracy: 0.9649 - val_loss: 7.6303 - val_accuracy: 0.1980 Epoch 13/20 129/129 [==============================] - 34s 267ms/step - loss: 0.1236 - accuracy: 0.9629 - val_loss: 7.8121 - val_accuracy: 0.1980 Epoch 14/20 129/129 [==============================] - 42s 323ms/step - loss: 0.1397 - accuracy: 0.9629 - val_loss: 7.4072 - val_accuracy: 0.2020 Epoch 15/20 129/129 [==============================] - 41s 316ms/step - loss: 0.1228 - accuracy: 0.9641 - val_loss: 8.3297 - val_accuracy: 0.1990 Epoch 16/20 129/129 [==============================] - 35s 270ms/step - loss: 0.1446 - accuracy: 0.9568 - val_loss: 7.9732 - val_accuracy: 0.1932 Epoch 17/20 129/129 [==============================] - 33s 257ms/step - loss: 0.1376 - accuracy: 0.9580 - val_loss: 7.3570 - val_accuracy: 0.2000 Epoch 18/20 129/129 [==============================] - 32s 251ms/step - loss: 0.1013 - accuracy: 0.9668 - val_loss: 7.8063 - val_accuracy: 0.1883 Epoch 19/20 129/129 [==============================] - 33s 253ms/step - loss: 0.0881 - accuracy: 0.9663 - val_loss: 8.1047 - val_accuracy: 0.1971 Epoch 20/20 129/129 [==============================] - 40s 314ms/step - loss: 0.0906 - accuracy: 0.9702 - val_loss: 8.2074 - val_accuracy: 0.1932
vis=history.history #gives entire trained DL model info like loss,validation loss and accuracy
vis
{'loss': [3.948608875274658,
3.787815809249878,
3.1219520568847656,
1.9319437742233276,
1.0306397676467896,
0.5366322994232178,
0.3896963894367218,
0.30352678894996643,
0.28283756971359253,
0.20524688065052032,
0.15468907356262207,
0.13382349908351898,
0.12357614189386368,
0.13970394432544708,
0.12281706184148788,
0.14461937546730042,
0.13757672905921936,
0.10125679522752762,
0.08813934028148651,
0.09059755504131317],
'accuracy': [0.02709299512207508,
0.07029534131288528,
0.22870393097400665,
0.49450817704200745,
0.7366365790367126,
0.8640468716621399,
0.898950457572937,
0.9226263165473938,
0.9297046661376953,
0.952160120010376,
0.9572858214378357,
0.9648523330688477,
0.9628996849060059,
0.9628996849060059,
0.964120090007782,
0.9567976593971252,
0.9580180644989014,
0.9668049812316895,
0.966316819190979,
0.9702221155166626],
'val_loss': [3.928774118423462,
3.6690032482147217,
3.522711753845215,
4.15183687210083,
4.910440921783447,
6.253957748413086,
6.855329513549805,
6.775447845458984,
7.059556484222412,
7.7634758949279785,
7.731235027313232,
7.630293846130371,
7.812135219573975,
7.407196521759033,
8.329679489135742,
7.973155975341797,
7.357020378112793,
7.8063201904296875,
8.104696273803711,
8.20743465423584],
'val_accuracy': [0.05170731619000435,
0.1034146323800087,
0.16878049075603485,
0.1921951174736023,
0.19707317650318146,
0.1921951174736023,
0.20000000298023224,
0.18731707334518433,
0.1912195086479187,
0.19609756767749786,
0.1921951174736023,
0.19804878532886505,
0.19804878532886505,
0.20195122063159943,
0.19902439415454865,
0.1931707262992859,
0.20000000298023224,
0.18829268217086792,
0.19707317650318146,
0.1931707262992859]}
#<----PERFORMANCE------>
#LOSS VIZULATIZATION
plt.figure(figsize=(5,5))
plt.plot(vis['loss'],color="red",label="loss")
#plt.plot(vis['val_loss'],color="blue",label="val_loss")
plt.show()
#ACCURACY VIZULIZATION
plt.figure(figsize=(5,5))
plt.plot(vis['accuracy'],color="green",label="accuracy")
#plt.plot(vis['val_accuracy'],color="pink",label="val_accuracy")
plt.show()
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
# Assuming you have already trained your model and obtained predictions
predictions = model.predict(train_images)
predicted_labels = np.argmax(predictions, axis=1)
# Calculate accuracy
accuracy = accuracy_score(train_labels, predicted_labels)
print("Accuracy:", accuracy*100,"%")
# Calculate precision
precision = precision_score(train_labels, predicted_labels, average='weighted')
print("Precision:", precision*100, "%")
# Calculate recall
recall = recall_score(train_labels, predicted_labels, average='weighted')
print("Recall:", recall*100, "%")
# Calculate F1 score
f1 = f1_score(train_labels, predicted_labels, average='weighted')
print("F1 Score:", f1*100,"%")
129/129 [==============================] - 18s 75ms/step Accuracy: 98.2182084452038 % Precision: 98.46826504037023 % Recall: 98.2182084452038 % F1 Score: 98.13702772110528 %
def load_data(folder):
images = []
labels = []
class_names = sorted(os.listdir(folder))
for class_index, class_name in enumerate(class_names):
class_path = os.path.join(folder, class_name)
for image_name in os.listdir(class_path):
image_path = os.path.join(class_path, image_name)
image = cv2.imread(image_path)
# Convert the color space from BGR to RGB
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
image = cv2.resize(image, (256,256))
images.append(image)
labels.append(class_index)
return images, labels
# Load and preprocess image data
test_images, test_labels = load_data('test_data')
# Convert lists to numpy arrays
test_images = np.array(test_images)
test_labels = np.array(test_labels)
# Normalize pixel values to the range [0, 1]
test_images = test_images.astype('float32') / 255.0
# Print shapes for verification
print("Test images shape:", test_images.shape)
print("Test labels shape:", test_labels.shape)
Test images shape: (520, 256, 256, 3) Test labels shape: (520,)
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
# Assuming you have already trained your model and obtained predictions
predictions = model.predict(test_images)
predicted_labels = np.argmax(predictions, axis=1)
# Calculate accuracy
accuracy = accuracy_score(test_labels, predicted_labels)
# Calculate precision
precision = precision_score(test_labels, predicted_labels, average='weighted')
# Calculate recall
recall = recall_score(test_labels, predicted_labels, average='weighted')
# Calculate F1 score
f1 = f1_score(test_labels, predicted_labels, average='weighted')
print("Accuracy:", accuracy*100,"%")
print("Precision:", precision*100, "%")
print("Recall:", recall*100, "%")
print("F1 Score:", f1*100,"%")
17/17 [==============================] - 3s 69ms/step Accuracy: 83.65384615384616 % Precision: 85.36195754313162 % Recall: 83.65384615384616 % F1 Score: 83.7713987154614 %
from numpy import *
images_list=[]
prediction_list=[]
def pred(image_name):
display=cv2.imread(image_name)
display = cv2.cvtColor(display, cv2.COLOR_BGR2RGB)
img=cv2.imread(image_name)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
resize=cv2.resize(img,(256,256))
print(resize.shape)
#optimize the new image
resize=resize/255
#expand your image array
img=expand_dims(resize,0)
predictions = model.predict(img)
# Convert the predicted probabilities to class labels
predicted_labels = np.argmax(predictions, axis=1)
print("The Sub Class detected in Breast Cancer is: ",classes[predicted_labels[0]])
images_list.append(display)
prediction_list.append(predicted_labels)
#loaded model prediticion
for i in range(1,5):
pred('pic'+str(i)+'.jpg')
(256, 256, 3) 1/1 [==============================] - 0s 36ms/step The Sub Class detected in Breast Cancer is: Africanized Honey Bees (Killer Bees) (256, 256, 3) 1/1 [==============================] - 0s 69ms/step The Sub Class detected in Breast Cancer is: Anaphe reticulata (256, 256, 3) 1/1 [==============================] - 0s 50ms/step The Sub Class detected in Breast Cancer is: Anomis sabulifera (256, 256, 3) 1/1 [==============================] - 0s 48ms/step The Sub Class detected in Breast Cancer is: Ants
fig,ax=plt.subplots(nrows=2,ncols=2,figsize=(20,20))
ax=ax.flatten()
for i,j in enumerate(images_list[:4]):
ax[i].imshow(j) #already in minimal scaled range
ax[i].title.set_text("Label: "+str(prediction_list[i][0])+"; The Sub Class detected in Breast Cancer is: "+str(classes[prediction_list[i][0]]))
model.save('pest.keras')